RESTAURANT
One operator we work with opened in Korea in 2022. Today they run four locations, three concepts, and one Michelin star. Here is the framework that made it possible.
One operator we work with opened in Korea in 2022. Today they run four locations. Three different concepts. One of them has a Michelin star. The thing that made that possible is the same thing most operators try to skip when they open location two. This is what to do instead.
The first restaurant works because you were standing in it. You wrote the recipe. You trained the line cook. You closed the bar at 1am for eighteen months. The standards live in your head and your hands.
Then you sign the lease on the second location. You hire a GM. You hand them a Google Drive with your manual. You walk through prep with the new line cook for two weeks. You open the doors.
A few months later the new GM is calling on Saturdays asking how the chana masala gets prepped. The line cook there is plating differently. Guests are saying it does not taste the same as the original. You cannot tell the new GM what is wrong because you cannot reconstruct any single shift. You just know something is off.
This is not a talent problem. It is a standards-transfer problem. And the difference between an operator who scales and one who stalls comes down to whether they treat standards as portable software or as oral tradition.
"Every great chef I had ever worked with was also a single point of failure. The industry calls that talent. We call it fragile." David Castro Zabarain, founder, Restaurant Codex.
When we talk to operators running two locations, five locations, ten locations, the same five words come up. Not in any order. Five categories of pain that map almost perfectly onto what the new GM does not have on day one.
Standards. People. Margins. Trust. Time.
The flagship runs because all five live in you. The new site does not feel like the flagship because all five have to be transferred, and nobody hands you a system for transferring them. Here is what each one looks like, and how the Korea operator handled it differently.
You wrote the SOPs. You shared the Google Drive. The new GM has access. So why does Tuesday at the new site look nothing like Tuesday at the flagship?
Because an SOP nobody opens is a wish. By the time a Friday night cook needs the steps for cleaning the fryer, they are not opening Google Drive. They are guessing. Or asking the closest person.
You shared a document. You did not share a standard.
The fix is making the SOP live where the work happens. On the staff phone, gated by role, with photos that prove the step was done. Restaurant Codex calls this My Manual and it sits in the staff member's pocket all shift. Each entry has photo verification, version control, and role-based access. Hosts see host SOPs. Line cooks see line SOPs.
Operators say "I can't find good people." The real pain is not finding them. The real pain is the new-hire churn before they are productive.
Here is what happens in practice. You hire a line cook on Monday. They shadow your most experienced cook. They pick up half of what they need. They figure out the rest by failing in front of guests. The good ones quit before they are productive. The bad ones stay and become the problem you cannot talk about.
At the new location, the most experienced cook is not actually that experienced. So shadowing transfers half the standards. And the one bilingual manager who translates between the kitchen and the front of house is now the single point of failure. When she is off, nothing holds.
The fix is making training finishable. A new hire should have a path, not a buddy. Role-based, multi-format, completion-tracked. Video where video helps. Written steps where reading helps. Real recipes from your library. Tests to verify they got it.
The bilingual problem also gets solved if your tools speak in any language. Codex Bot answers staff questions in the language the question was asked in. Spanish in, Spanish out. Korean in, Korean out. The one bilingual manager is no longer the bottleneck.
The recipe costs out at one number. The shift hits a different one. The next shift hits a third one. Your accountant cannot find the leak because the leak is not in your cost data. It is in execution.
At the flagship, you eyeballed it. You knew what the chili should look like. You corrected the cook on the line. Standards lived in your hands.
At a new location, nobody has that eye yet. The cook on Friday is eyeballing. The cook on Saturday is double-dosing. The recipe on paper does not match the recipe on the plate.
The fix is photos and weights, not better cost reports. Codex Recipes stores plate recipes and batch recipes separately. Every recipe carries a photo so the cook sees what the plate should look like before they make it. Allergen tags are built in. And the recipes live inside Codex Bot, so when a new hire asks how to prep a dish, they get back your recipe with photo and ingredient list.
The food cost gap is not closed by spreadsheets. It is closed by making the plate match the sheet, every shift, every cook.
This one is the killer. When something goes wrong on a Saturday at the new location, by Monday morning nobody remembers exactly what happened. You piece together a story from texts and half-remembered conversations. By the time you have the story, it is too late to do anything except generalize.
The new GM thinks you are interrogating. You think you are coaching. Neither of you can be specific because the evidence is gone.
Paper checklists make this worse. They end up complete because the cook signed them in twenty seconds at end of shift. They prove nothing.
The fix is photo verification on every task that matters. The opening checklist is not a list. It is a contract between you and the team that the kitchen will be ready at 11am. Codex Tasks requires photos on temperature logs, line cleaning, allergen reset, and any other step where execution matters. Codex Activity turns every completion into a timestamped record. By Monday morning you do not have to remember. You scroll the feed.
Trust by data, not gut. Coach by photo, not memory.
The cruelest version of the second-location problem. You can feel that the new site needs you. You can feel that the flagship will slip if you leave. So you do both, badly, and you have not taken a real day off in two years.
This breaks in a specific way. Half your phone time is answering the same handful of questions. How much chili. Where is the checklist. What do we do with this allergy. Every answer takes a minute. Every shift, ten of them. Across a year, that is your weekends and your sanity.
The fix is taking yourself out of the loop. Not by hiring more managers. More managers makes the standards-transfer problem worse. By making your manual answer the questions your staff are texting you.
This is the wedge in Codex Bot. It is trained on your own recipes, manual, and SOPs. A line cook asks in plain language. The bot returns your answer with photos. Your content. Your photos. Your standards.
The Korea operator did not skip any of the five. That is the whole story.
They opened location one in 2022 with the standards already built to be portable. The SOPs lived on the staff phone. The recipes carried photos. Training was role-based, not shadow-based. Tasks were photo-verified. The shift was legible.
So when they opened location two, they were not transferring oral tradition into a new GM's head. They were cloning a system that already worked. Same with location three. Same with location four. Three of those locations run different concepts. The concepts are different. The operating layer underneath is the same.
One of them earned a Michelin star along the way.
Another operator we work with opened in Mexico City in June 2024. By December 2025 they were opening location two. Same pattern. The systems they installed at location one made the second opening a clone-and-customize operation, not a fresh build.
This is what changes when standards are portable. Opening the next location is not a new project. It is the same project, deployed again.
If you are about to open a second location, or you just opened one and it is not landing right, here is what to do this week.
The second location does not have to be a rebuild of the first. Most of the gap is because the system to transfer standards never existed at your operation. The fix is not heroics. The fix is software that holds your standards portably, so opening the next location is closer to a copy-paste than to a rebuild.
Every operator we work with is profitable. All of them running double-digit margins. We open the live product and walk you through how the five-word framework gets installed shift to shift, site to site. Real product, no slides. You can book a walkthrough using the link below.
